Follow
Quite impressive work on LLM interpretability, with applications.
https://transformer-circuits.pub/2024/scaling-monosemanticity/index.html
